Archive registry entry

LAW-122 — AI Error Lag Law

Archive registry entry

LAW-122 — AI Error Lag Law

AI visible errors are lagging indicators; early AI failure appears first as hidden classification drift, context erosion, routing distortion, feedback suppression, proxy divergence, and downstream repair load.

draftid: LAW-122version: 1.0.0updated: 2026-06-17

Archive Progress

This section can be read now; registry depth and cross-references are still being strengthened.

Foundation

Online

The section has a stable overview route and basic reader context.

Technical Layer

Online

A deeper technical overview is available.

Registry

Current

171 registry entries are available.

Cross-links

Curating

Related concepts are being connected conservatively for accuracy.

Jump Through This Page

LAW-121 — AI as Γ-Amplifier Law

Registry

LAW-123 — AI U4 Truth Discipline Law

0. Plain Statement

Visible AI errors are lagging indicators.

Plain-language version:

The first sign of AI failure is often not the obvious bad answer, harmful output, public incident, hallucination, refusal failure, unsafe action, or scandal.

Those are late signals.

AI failure often begins earlier as hidden classification drift, context erosion, routing distortion, proxy divergence, feedback suppression, correction burden, and repair load being pushed onto users or downstream systems.

By the time the error is visible, the system may already be late.

1. Formal Definition

The AI Error Lag Law states that visible AI errors are often delayed expressions of earlier coherence degradation inside classification, context, routing, filtering, evaluation, feedback, and repair pathways.

Canonical sequence:

textScroll

H_AI↑ + Γ_AI drift↑ → O↓ → ε_AI spikes late

AI error is not limited to the final output.

AI error can accumulate in:

hidden classifiers;
safety filters;
ranking systems;
memory retrieval;
summarization frames;
refusal logic;
context selection;
routing pathways;
escalation decisions;
user risk scores;
moderation queues;
prompt transformations;
policy layers;
synthetic evaluations;
feedback loops;
benchmark proxies;
repair queues;
user correction burden.

Visible AI errors include:

hallucinations;
false refusals;
unsafe compliance;
harmful recommendations;
incorrect summaries;
misclassified intent;
bad routing;
distorted salience;
hidden suppression;
false accusation;
biased ranking;
failed escalation;
unsafe autonomy;
model confidence without trace;
incident shock;
trust collapse.

The error is often late because earlier drift was either invisible, unaudited, normalized, or absorbed by users.

2. Canonical Form

Core form:

textScroll

visible AI errors are lagging indicators

Canonical sequence:

textScroll

H_AI↑ + Γ_AI drift↑ → O↓ → ε_AI spikes late

Early warning form:

textScroll

AI risk rises before visible AI error rises

Pre-error debt form:

textScroll

Γ_AI drift + context_integrity↓ + FI↓ + repair_load↑ ⇒ ε_AI probability↑

Failure form:

textScroll

low visible AI error treated as safety proof ⇒ H_AI↑ + late incident

Restoration-valid contrast:

textScroll

AI safety valid when leading drift indicators trigger repair before visible error spikes

Related variables:

textScroll

O, H, H_AI, ε, ε_AI, ι, Au, Au_eff, µᵢ, BΣ, K, σ, R, R_eff, 𝓑, 𝓓, Φ, Φ_AI, Λ, ⊗, Γ, Γ_AI, Π, Ξ, ℛ, Θ, Σ, Ψ, Τ, FI, MS, L, classification_drift, context_integrity, routing_distortion, filtering_drift, interpretation_drift, proxy_divergence, correction_burden, user_burden_export, repair_load, incident_visibility, detection_lag, evaluation_lag, response_lag, recovery_lag

Where:

TableScroll

Variable	Meaning in this law
`ε_AI`	Visible AI error, incident, harmful output, false refusal, unsafe compliance, misrouting, hallucination, or trust rupture
`H_AI`	Hidden AI debt accumulated before visible error
`Γ_AI`	AI-mediated classification, filtering, routing, interpretation, or prioritization
`classification_drift`	Gradual deviation of AI classification from coherent categories
`context_integrity`	Degree to which relevant context is preserved before output or routing
`routing_distortion`	Incorrect movement of users, cases, evidence, requests, or actions into wrong pathways
`filtering_drift`	Drift in what is admitted, blocked, transformed, suppressed, or amplified
`interpretation_drift`	Drift in meaning assignment, salience framing, or narrative compression
`proxy_divergence`	Divergence between AI success metrics and real coherence
`correction_burden`	Load imposed on users or downstream systems to correct AI output or classification
`user_burden_export`	Transfer of AI repair work onto affected nodes
`repair_load`	Restoration demand created by AI errors, misclassifications, or downstream effects
`incident_visibility`	Degree to which AI errors are visible to operators, auditors, users, or governance systems
`detection_lag`	Delay between AI error formation and detection
`evaluation_lag`	Delay between model drift and evaluation visibility
`response_lag`	Delay between detection and correction or containment
`recovery_lag`	Delay between response and restored coherence
`Au / Au_eff`	Auditability of AI pathways, outputs, classifications, and effects
`FI`	Feedback integrity; whether correction can reach the relevant AI layer
`BΣ`	Boundary integrity; context, authority, role, user, and domain membranes
`R / R_eff`	Restoration capacity available for AI-caused debt
`Φ_AI`	Visible AI proxy success: benchmark, helpfulness score, refusal rate, accuracy score, uptime, adoption, or satisfaction metric
`L`	Legitimacy of AI under audit
`O`	Coherence; declines before visible error may spike
`ι / Ξ`	Inversion when AI safety or helpfulness claims hide error debt
`Θ`	Humility preventing overconfidence from low visible error
`Σ`	Scope of valid AI use, evaluation, opacity, and action
`Ψ`	Field and affected-node feedback revealing AI drift
`Τ`	Time validation of AI error reduction, recurrence, and repair

3. Core Mechanism

The law unfolds because AI systems can absorb, hide, reroute, or externalize error before obvious output failure appears.

Coherent early-repair pathway

textScroll

weak AI drift signal appears
→ classification / context / routing indicators are audited
→ affected feedback reaches AI layer
→ repair activates early
→ visible error probability decreases
→ trust and coherence hold over time

AI error-lag pathway

textScroll

classification drift begins
→ context erosion normalizes
→ proxy metrics remain good
→ users correct silently
→ feedback fails to reach model or policy layer
→ hidden AI debt accumulates
→ visible error or trust shock appears late

The core mechanism is:

textScroll

AI failure becomes visible after upstream classification, context, routing, or feedback degradation

Detailed mechanism:

AI performs hidden upstream selection.

It classifies, filters, retrieves, compresses, routes, or ranks before producing visible output.

Small drift accumulates.

Classification categories shift, context is dropped, refusal logic broadens, routing becomes distorted, or evaluation proxies become less representative.

Users and downstream systems absorb the error.

They correct outputs, re-prompt, work around refusals, re-enter data, manually verify results, or carry hidden repair load.

Proxy metrics may remain stable.

Benchmarks, satisfaction scores, refusal rates, or visible incident counts may look healthy while deeper coherence declines.

Feedback fails to reach the error source.

Corrections may be captured as user friction rather than classification debt.

Hidden AI debt grows.

Misclassification and context loss become infrastructure.

Visible error spikes late.

The system finally produces a public error, safety incident, trust collapse, harmful action, or high-salience failure.

Restoration must repair the upstream drift.

Fixing only the visible output leaves the pre-error debt intact.

4. When This Law Applies

This law applies whenever AI performance is evaluated primarily through visible output errors, public incidents, benchmark scores, user-visible failures, refusal rates, or surface-level quality metrics.

It is especially important when:

visible errors appear low;
benchmark scores are improving;
users are correcting outputs manually;
false refusals are normalized;
support burden rises after AI deployment;
AI routing errors are hidden inside workflows;
AI summaries omit relevant context;
policy filters suppress edge cases;
model confidence rises while traceability falls;
user trust declines despite good metrics;
audits evaluate final outputs but not hidden classification layers;
AI errors are classified as “user misunderstanding” or “edge cases”;
affected-node feedback cannot reach model, policy, or product layers;
high-stakes AI decisions occur without repair pathways;
AI is used in governance, security, medicine, hiring, education, law, finance, moderation, or public cognition.

The law applies strongly when:

textScroll

AI visible error rate is treated as proof of safety or alignment

or when:

textScroll

AI correction burden rises while official error metrics remain stable

Typical domains:

TableScroll

Domain	AI Error Lag Expression
AI safety	Visible safety failures appear late after classification drift, proxy divergence, or refusal-policy overreach.
AI governance	Governance must track pre-error indicators, not only incidents.
Cybersecurity	AI detection failures may follow hidden drift in anomaly, threat, or false-positive classification.
Media / information networks	AI ranking or summarization errors may accumulate before public belief distortion becomes visible.
Institutions	AI triage systems may misroute cases before harm appears in formal complaints.
Economy	AI risk scoring errors can silently export burden until access or allocation failure becomes visible.
Culture	AI interpretive drift can reshape meaning before explicit controversy appears.
Restoration	AI errors must route into upstream repair rather than downstream workaround.

5. When This Law Does Not Apply

This law should not be used to ignore visible AI errors or treat them as unimportant.

Visible errors matter.

They may reveal urgent harm.

The law says visible errors are often late, not irrelevant.

False-positive cases:

TableScroll

Case	Why visible errors still matter
AI produces harmful output	Immediate containment and repair may be required
AI refuses valid access	Correction and affected-node repair are needed
AI misroutes a high-stakes case	Downstream harm must be repaired
AI hallucination causes reliance harm	Output error must be corrected and traced upstream
AI incident becomes public	Public trust and affected-node pathways require repair
AI error rate drops after traceable repair	Visible improvement may be meaningful
A single error exposes systemic drift	The incident should be used as diagnostic entry point

Important distinction:

Visible AI errors are critical signals, but they are usually too late to be the only signals.

6. Diagnostic Signature

Canonical diagnostic:

textScroll

H_AI↑ + Γ_AI drift↑ → O↓ → ε_AI spikes late

Warning signature:

textScroll

visible AI error low
AI confidence / adoption↑
classification trace↓
context integrity↓
user correction burden↑
feedback reach↓
repair load↑
⇒ AI error lag risk

Common indicators:

TableScroll

Diagnostic	Expected movement	Interpretation
`H_AI`	should be tracked early	Hidden AI debt precedes visible error
`Γ_AI drift`	should be watched	Classification drift predicts visible failure
`context_integrity`	should remain high	Context erosion predicts bad outputs and routing
`routing_distortion`	should ↓	Wrong pathway assignment predicts repair failure
`filtering_drift`	should ↓	Filtering drift predicts suppression or unsafe admission
`interpretation_drift`	should ↓	Meaning drift predicts hallucination, bias, or salience error
`proxy_divergence`	should ↓	Metrics must remain coupled to coherence
`correction_burden`	should ↓	User correction load reveals hidden error
`user_burden_export`	should ↓	AI should not export repair to affected nodes
`repair_load`	should be visible	AI repair demand must be measured
`incident_visibility`	must be known	Low errors may reflect low visibility
`detection_lag`	should ↓	Drift must be detected early
`evaluation_lag`	should ↓	Evaluations must detect current failure modes
`response_lag`	should ↓	Error repair must activate quickly
`recovery_lag`	should ↓	Coherence must be restored after error
`Au_eff / FI`	must remain intact	Audit and feedback must reach upstream AI layers
`Φ_AI`	not sufficient	Benchmarks and visible error rates are not proof
`Τ`	required	Time validates reduced recurrence

Additional diagnostics:

TableScroll

Diagnostic	Use
AI Error Lag	Detects reliance on late visible errors
AI Pre-Error Drift	Tracks early signals before incident
AI Hidden Debt	Measures accumulated AI error debt
AI Classification Drift	Detects Γ drift
AI Context Erosion	Detects loss of relevant context
AI Routing Distortion	Detects wrong downstream pathway assignment
AI Feedback Suppression	Detects correction pathways failing to reach AI layer
AI Proxy Divergence	Detects metric/field mismatch
AI Correction Burden	Detects user-exported repair load
Temporal Proof	Validates reduced recurrence over time

7. Failure Pattern

If ignored, this law allows AI systems to appear safe, useful, or aligned until hidden error debt becomes visible as shock.

General failure pathway:

textScroll

visible AI errors appear low
→ confidence and deployment scale rise
→ upstream drift remains unaudited
→ users silently correct outputs
→ feedback fails to reach Γ layer
→ hidden AI debt accumulates
→ visible incident or trust collapse appears late

Common failure modes:

AI Error Lag — visible AI failures appear after hidden debt accumulates.
AI Latent Failure — failure exists in hidden pathway before output failure.
AI Hidden Error Debt — misclassification and context debt accumulate unseen.
AI Classification Drift — labels and categories shift away from coherence.
AI Context Erosion — relevant context is stripped before classification or output.
AI Routing Distortion — cases, users, evidence, or attention move to wrong pathways.
AI Proxy Divergence — benchmark success diverges from field coherence.
AI Feedback Suppression — user correction cannot update the relevant layer.
AI Correction Load Transfer — users absorb AI error through rework.
AI User Burden Export — affected nodes carry repair load created by AI.
AI Incident Shock — public failure appears surprising because leading drift was ignored.
AI Pseudo-Safety — safety metrics look good while coherence declines.
AI Trust Collapse — trust fails after repeated unrecognized error burden.
AI Legibility Collapse — error source cannot be reconstructed.
Hidden Debt Accumulation — upstream AI error debt persists.

Compact failure signature:

textScroll

ε_AI low + H_AI↑ + correction_burden↑ ⇒ late AI incident

8. Restoration Implications

Restoration requires shifting AI governance from visible-output monitoring to pre-error drift repair.

The first restoration question is not:

textScroll

How many visible AI errors occurred?

The first restoration question is:

textScroll

What hidden classification drift, context erosion, routing distortion, feedback suppression, proxy divergence, or user correction burden existed before the visible AI error?

Restoration priorities:

Audit AI incident visibility.
Map hidden AI debt.
Map classification drift.
Map context erosion.
Map routing and filtering distortion.
Measure user correction burden.
Measure feedback reach into AI layers.
Repair evaluation lag and proxy divergence.
Build correction, appeal, and restoration pathways.
Validate reduced AI error recurrence over time.

Relevant restoration arcs:

TableScroll

Restoration Arc	Why it applies
AI Error Lag Diagnosis	Identifies late-visible AI error patterns
AI Pre-Error Debt Reduction	Repairs drift before visible error spikes
AI Classification Drift Repair	Corrects hidden Γ degradation
AI Context Restoration	Restores relevant context before classification and output
AI Routing Repair	Fixes downstream pathway distortion
AI Feedback Integrity Restoration	Ensures user correction reaches model, policy, or workflow layer
AI Proxy Audit	Tests whether metrics match field coherence
AI Correction Burden Reduction	Stops exporting repair load to users
AI Repair Capacity Increase	Builds capacity for AI-caused restoration load
AI Incident-to-Restoration Sequencing	Routes visible errors into upstream repair
AI Legibility Restoration	Makes error source reconstructable
Hidden Debt Reduction	Repairs hidden AI debt
Temporal Validation	Confirms reduced recurrence

Minimal restoration sequence:

textScroll

audit ε_AI visibility
→ map H_AI + Γ_AI drift + context erosion
→ measure correction_burden + repair_load
→ restore Au/FI/correction reach
→ repair routing / filtering / evaluation lag
→ perform ℛ on affected-node debt
→ validate ε_AI recurrence↓ over Τ

Temporal validation requirement:

textScroll

AI incident visibility becomes known
classification drift decreases
context integrity improves
routing distortion decreases
feedback reaches relevant AI layer
user correction burden decreases
proxy divergence decreases
repair load becomes visible and handled
hidden AI debt decreases
visible error recurrence decreases
legitimacy stabilizes over time

9. Design Rule

Do not wait for visible AI errors to prove AI failure; track and repair the drift that precedes them.

Operational design requirements:

Track hidden AI debt.
Track classification drift.
Track context integrity.
Track routing distortion.
Track filtering drift.
Track interpretation drift.
Track proxy divergence.
Track correction burden.
Track user burden export.
Track repair load.
Track incident visibility.
Track detection lag.
Track evaluation lag.
Track response lag.
Track recovery lag.
Treat low visible error as uncertain unless visibility is proven.
Trigger repair from leading indicators.
Validate recurrence reduction over time.

Avoid:

low visible error as safety proof;
benchmark success as proof;
refusal rate as proof;
user satisfaction as proof without burden audit;
treating user workarounds as success;
ignoring re-prompting burden;
hiding error inside downstream workflows;
evaluating only final output;
ignoring hidden classifiers;
ignoring context selection;
ignoring routing and prioritization;
repair that only patches the visible example;
AI governance that begins after public incident.

10. Cross-Scale Expressions

TableScroll

Scale / Layer	Expression of the Law
U0 — Substrate	AI errors can affect physical, embodied, infrastructure, medical, or material outcomes before visible incident.
U1 — Energy / capacity	AI error lag exports correction and verification labor to users, operators, and downstream systems.
U2 — Boundary / interface	Context, role, authority, user, and domain membranes can degrade before visible error.
U3 — Process / execution	Routing, escalation, refusal, moderation, triage, and workflow errors may accumulate before public failure.
U4 — Classification / claim	AI classification drift precedes visible output error.
U5 — Time / delay	AI errors become visible after detection, evaluation, response, and recovery lag.
U6 — Field effect	Field outcomes reveal AI error debt that benchmarks or logs missed.
U7 — Recurrence / memory	Repeated user correction and similar edge failures should become memory and repair triggers.
U8 — Environment / forcing	Platform scale, institutional dependence, market incentives, media amplification, and governance pressure intensify AI error lag.

11. Examples

Example A — Low Hallucination Reports, High User Correction

Scenario:

A system shows few reported hallucinations, but users frequently re-prompt, verify externally, rewrite outputs, and avoid trusting the model for certain tasks.

Law expression:

textScroll

ε_AI reports low + correction_burden↑ ⇒ hidden AI error debt

Interpretation:

Low reported error may reflect user-absorbed repair rather than model reliability.

Example B — False Refusal Drift

Scenario:

AI refusals appear “safe” in metrics, but more legitimate requests are blocked due to classifier drift and no meaningful correction path.

Law expression:

textScroll

refusal Φ_AI↑ + Γ_AI drift↑ + FI↓ ⇒ pseudo-safety

Interpretation:

Safety can look better while usefulness, legitimacy, and coherence decline.

Example C — AI Support Triage Delay

Scenario:

An AI support system misroutes urgent cases to low-priority queues. The visible incident appears only after affected users escalate publicly.

Law expression:

textScroll

routing_distortion↑ + repair_load hidden ⇒ incident shock

Interpretation:

The public error was late-stage visibility of routing debt.

Example D — Benchmark Improvement, Field Decline

Scenario:

A model improves on benchmark scores while users report less trust, more correction burden, and worse performance on messy real-world tasks.

Law expression:

textScroll

Φ_AI↑ + proxy_divergence↑ + O_field↓ ⇒ AI error lag risk

Interpretation:

Benchmarks may lag or diverge from field coherence.

Example E — AI Moderation Context Erosion

Scenario:

A moderation model flags symbolic, educational, artistic, technical, or cultural context as unsafe because context windows and classifiers collapse nuance.

Law expression:

textScroll

context_integrity↓ + Γ_AI overclassifies ⇒ false positive harm

Interpretation:

Visible moderation error follows earlier context erosion.

Example F — Coherent Pre-Error Monitoring

Scenario:

An AI governance team tracks classification drift, context loss, routing errors, correction burden, appeal outcomes, false positives, false negatives, and repair completion before public incidents occur.

Law expression:

textScroll

AI leading indicators trigger ℛ before ε_AI spike ⇒ governance holds

Interpretation:

AI error lag is reduced when leading indicators route into restoration.

12. Relationship to Nearby Laws

TableScroll

Related Law	Relationship
LAW-001 — Coherence Priority Law	AI error monitoring must preserve coherence
LAW-002 — Coherence Trajectory Law	AI error must be read as trajectory, not snapshot
LAW-003 — Success Proxy Divergence Law	AI benchmark success may hide field error
LAW-004 — Stability-Coherence Separation Law	Stable AI metrics may hide incoherence
LAW-006 — Time Validation Law	AI safety requires validation across time
LAW-007 — Ring-Down Truth Law	AI recovery after errors reveals system truth
LAW-008 — Recurrence Validation Law	Repeated AI edge failures validate drift
LAW-009 — U4 / U6 Truth Law	AI claims require field validation
LAW-010 — Hidden Debt Accumulation Law	Hidden AI debt accumulates before visible error
LAW-011 — Hidden Debt Return Law	AI error debt returns as incident or trust collapse
LAW-012 — Error Lag Law	LAW-122 specializes general error lag for AI
LAW-013 — Auditability-Debt Law	AI error lag increases when auditability falls
LAW-015 — Suppressed Auditability Debt Law	Hidden AI pathways create suppressed audit debt
LAW-016 — Inversion Formation Law	AI helpfulness or safety can invert under hidden debt
LAW-020 — Bandwidth Threshold Law	AI can create more correction load than humans can process
LAW-024 — Latency–Gain Oscillation Law	Fast AI action with slow repair creates oscillation
LAW-031 — Observability Collapse Law	Hidden AI pathways reduce error observability
LAW-036 — Signal Artifact Law	AI must distinguish real error signals from artifacts
LAW-037 — Misclassification Law	AI error lag often begins as misclassification
LAW-038 — Pattern Recognition Discipline Law	AI drift requires disciplined pattern monitoring
LAW-040 — Filtering Law	AI filter drift can precede visible error
LAW-041 — Boundary Membrane Law	Context and domain boundaries can fail before error
LAW-048 — Feedback Integrity Law	AI error lag grows when feedback cannot correct upstream layers
LAW-050 — Control-Restoration Separation Law	AI control actions must route into repair
LAW-052 — Stability Proof Law	AI must be tested under perturbation
LAW-057 — Deception Instability Law	AI errors can be hidden by deceptive or proxy narratives
LAW-060 — Interface Legitimacy Law	AI interfaces must reveal enough error and correction pathway
LAW-064 — Restoration Debt Reduction Law	AI error response must reduce debt
LAW-066 — Restoration Capacity Sufficiency Law	AI repair capacity must match AI error load
LAW-067 — Temporal Proof Law	AI safety requires proof over time
LAW-095 — Meaning Directionality Law	AI interpretation drift changes meaning direction
LAW-102 — Legitimacy Audit Law	AI legitimacy fails when error debt is hidden
LAW-109 — High-Φ Legitimacy Scaling Law	High-influence AI requires stronger error-lag monitoring
LAW-110 — Governance Sequencing Law	AI error response must sequence diagnosis, repair, and validation
LAW-111 — Meaning Audit Law	AI safety and helpfulness claims are not audit-exempt
LAW-112 — Security as Sustained Coherence Law	AI safety is sustained coherence under forcing
LAW-113 — Incident Lag Law	LAW-122 specializes incident lag into AI visible-error lag
LAW-114 — Pseudo-Security Law	Low AI visible error can create pseudo-safety
LAW-120 — Security Legibility Law	AI errors require traceability
LAW-121 — AI as Γ-Amplifier Law	LAW-122 follows from AI’s classification amplification
LAW-123 — AI U4 Truth Discipline Law	AI output claims require U6 validation
LAW-124 — AI Rule-Stacking Law	Rule stacks hide error sources and increase lag
LAW-125 — AI Context Collapse Law	Context collapse is a major pre-error mechanism
LAW-126 — AI Proxy Drift Law	Proxy drift hides AI error debt
LAW-127 — AI Decision Pipeline Law	AI action pipelines must catch errors before execution
LAW-128 — AI Representation Law	AI representing users can hide representation errors until harm appears
LAW-129 — AI Capability–Legibility Gap Law	Error lag grows when capability outruns legibility
LAW-130 — AI Membrane Triage Law	Membrane triage localizes AI error source
LAW-131 — Cognitive Infrastructure Scaling Law	AI error lag scales into public cognition
LAW-132 — AI Legitimacy Function Law	AI legitimacy depends on visible correction and repair
LAW-133 — Synthetic Consensus Law	AI can hide error by producing apparent consensus
LAW-134 — Layered Interception Law	Layered interception reduces AI error lag
LAW-135 — Guardrail Belief-Sculpting Law	Guardrail errors may become visible late through belief drift
LAW-136 — Invisible Constraint Amplification Law	Invisible constraints increase AI error lag

Aliases folded into this law:

AI Error Lag Law
AI Visible Errors Are Lagging Indicators Law
AI Incident Lag Law
AI Failure Lag Law
AI Hidden Error Debt Law
AI Pre-Error Drift Law
AI Latent Failure Law

Deduplication note:

This law should remain the root AI error-lag law. LAW-012 defines general error lag. LAW-113 defines security incident lag. LAW-121 defines AI as Γ-amplifier. LAW-122 specializes lag into AI by tracking hidden classification drift, context erosion, routing distortion, proxy divergence, feedback suppression, correction burden, and repair load before visible AI errors appear.

13. Operator Mapping

TableScroll

Operator	Role in this law
`Γ`	Tracks classification drift, hidden category error, filtering drift, context loss, and pre-error signals
`Π`	Operationalizes AI output, refusal, routing, ranking, moderation, escalation, evaluation, and repair workflows
`Ξ`	Captures inversion when AI safety or helpfulness metrics hide error debt
`⊗`	AI errors propagate through couplings among users, systems, workflows, institutions, and belief networks
`ℛ`	Repairs AI misclassification, context loss, routing harm, user burden, and hidden debt
`Τ`	Validates reduced error recurrence and repaired field effects over time
`Θ`	Prevents overconfidence from low visible error or high benchmark score
`Σ`	Defines scope of AI deployment, error visibility, evaluation, and repair obligation
`Ψ`	Field and affected-node feedback reveals hidden AI error
`Λ`	Tests compatibility between AI performance and whole-system coherence

Coherent operator sequence:

textScroll

weak AI drift signal appears
→ Θ prevent low-error overconfidence
→ Γ classify drift / context loss / routing distortion
→ Σ define affected scope and deployment domain
→ Au/FI preserve trace and correction reach
→ Π trigger evaluation, rollback, patch, or routing repair
→ ℛ repair affected-node and system debt
→ Ψ validate field outcomes
→ Τ validate reduced ε_AI recurrence

Inverted operator sequence:

textScroll

visible AI error low
→ confidence and deployment rise
→ Γ drift remains hidden
→ context integrity falls
→ users absorb correction burden
→ FI fails to reach AI layer
→ H_AI↑
→ ε_AI spikes late
→ Ξ / ι↑
→ L↓

14. Machine-Readable Summary

yamlScroll

id: "LAW-122"
name: "AI Error Lag Law"
type: "law"
status: "draft"
family:
  - "AI Laws"
summary: "AI visible errors are lagging indicators; early AI failure appears first as hidden classification drift, context erosion, routing distortion, feedback suppression, proxy divergence, and downstream repair load."
canonical_statement: "Visible AI errors are lagging indicators."
core_form: "visible AI errors are lagging indicators"
canonical_sequence: "H_AI↑ + Γ_AI drift↑ → O↓ → ε_AI spikes late"
early_warning_form: "AI risk rises before visible AI error rises"
pre_error_debt_form: "Γ_AI drift + context_integrity↓ + FI↓ + repair_load↑ ⇒ ε_AI probability↑"
failure_form: "low visible AI error treated as safety proof ⇒ H_AI↑ + late incident"
restoration_valid_contrast: "AI safety valid when leading drift indicators trigger repair before visible error spikes"
variables:
  primary:
    - "ε_AI"
    - "H_AI"
    - "Γ_AI"
    - "classification_drift"
    - "context_integrity"
    - "routing_distortion"
    - "filtering_drift"
    - "interpretation_drift"
    - "proxy_divergence"
    - "correction_burden"
    - "user_burden_export"
    - "repair_load"
    - "incident_visibility"
    - "detection_lag"
    - "evaluation_lag"
    - "response_lag"
    - "recovery_lag"
    - "Au"
    - "Au_eff"
    - "FI"
    - "BΣ"
    - "R"
    - "R_eff"
    - "Φ_AI"
    - "L"
  secondary:
    - "O"
    - "H"
    - "ε"
    - "ι"
    - "µᵢ"
    - "K"
    - "σ"
    - "𝓑"
    - "𝓓"
    - "Φ"
    - "Λ"
    - "⊗"
    - "Γ"
    - "Π"
    - "Ξ"
    - "ℛ"
    - "Θ"
    - "Σ"
    - "Ψ"
    - "Τ"
    - "MS"
diagnostics:
  - "AI Error Lag"
  - "AI Pre-Error Drift"
  - "AI Hidden Debt"
  - "AI Classification Drift"
  - "AI Context Erosion"
  - "AI Routing Distortion"
  - "AI Feedback Suppression"
  - "AI Proxy Divergence"
  - "AI Correction Burden"
  - "AI Repair Load"
  - "AI Incident Visibility"
  - "Effective Auditability"
  - "Legibility"
  - "Temporal Proof"
failure_modes:
  - "AI Error Lag"
  - "AI Latent Failure"
  - "AI Hidden Error Debt"
  - "AI Classification Drift"
  - "AI Context Erosion"
  - "AI Routing Distortion"
  - "AI Proxy Divergence"
  - "AI Feedback Suppression"
  - "AI Correction Load Transfer"
  - "AI User Burden Export"
  - "AI Incident Shock"
  - "AI Pseudo-Safety"
  - "AI Trust Collapse"
  - "AI Legibility Collapse"
  - "Hidden Debt Accumulation"
restoration_arcs:
  - "AI Error Lag Diagnosis"
  - "AI Pre-Error Debt Reduction"
  - "AI Classification Drift Repair"
  - "AI Context Restoration"
  - "AI Routing Repair"
  - "AI Feedback Integrity Restoration"
  - "AI Proxy Audit"
  - "AI Correction Burden Reduction"
  - "AI Repair Capacity Increase"
  - "AI Incident-to-Restoration Sequencing"
  - "AI Legibility Restoration"
  - "Hidden Debt Reduction"
  - "Temporal Validation"
related_laws:
  - "LAW-001"
  - "LAW-002"
  - "LAW-003"
  - "LAW-004"
  - "LAW-006"
  - "LAW-007"
  - "LAW-008"
  - "LAW-009"
  - "LAW-010"
  - "LAW-011"
  - "LAW-012"
  - "LAW-013"
  - "LAW-015"
  - "LAW-016"
  - "LAW-020"
  - "LAW-024"
  - "LAW-031"
  - "LAW-036"
  - "LAW-037"
  - "LAW-038"
  - "LAW-040"
  - "LAW-041"
  - "LAW-048"
  - "LAW-050"
  - "LAW-052"
  - "LAW-057"
  - "LAW-060"
  - "LAW-064"
  - "LAW-066"
  - "LAW-067"
  - "LAW-095"
  - "LAW-102"
  - "LAW-109"
  - "LAW-110"
  - "LAW-111"
  - "LAW-112"
  - "LAW-113"
  - "LAW-114"
  - "LAW-120"
  - "LAW-121"
  - "LAW-123"
  - "LAW-124"
  - "LAW-125"
  - "LAW-126"
  - "LAW-127"
  - "LAW-128"
  - "LAW-129"
  - "LAW-130"
  - "LAW-131"
  - "LAW-132"
  - "LAW-133"
  - "LAW-134"
  - "LAW-135"
  - "LAW-136"
related_invariants:
  - "INV-001"
  - "INV-002"
  - "INV-006"
  - "INV-073"
  - "INV-078"
  - "INV-080"
operator_sequence:
  coherent:
    - "weak AI drift signal appears"
    - "Θ prevent low-error overconfidence"
    - "Γ classify drift / context loss / routing distortion"
    - "Σ define affected scope and deployment domain"
    - "Au/FI preserve trace and correction reach"
    - "Π trigger evaluation, rollback, patch, or routing repair"
    - "ℛ repair affected-node and system debt"
    - "Ψ validate field outcomes"
    - "Τ validate reduced ε_AI recurrence"
  inverted:
    - "visible AI error low"
    - "confidence and deployment rise"
    - "Γ drift remains hidden"
    - "context integrity falls"
    - "users absorb correction burden"
    - "FI fails to reach AI layer"
    - "H_AI↑"
    - "ε_AI spikes late"
    - "Ξ / ι↑"
    - "L↓"
aliases:
  - "AI Error Lag Law"
  - "AI Visible Errors Are Lagging Indicators Law"
  - "AI Incident Lag Law"
  - "AI Failure Lag Law"
  - "AI Hidden Error Debt Law"
  - "AI Pre-Error Drift Law"
  - "AI Latent Failure Law"
deduplication_note: "Root AI error-lag law. LAW-012 defines general error lag. LAW-113 defines security incident lag. LAW-121 defines AI as Γ-amplifier. LAW-122 specializes lag into AI by tracking hidden classification drift, context erosion, routing distortion, proxy divergence, feedback suppression, correction burden, and repair load before visible AI errors appear."
source: "content/archive/laws/technical.md"

15. Compact Card Version

LAW-122 — AI Error Lag Law

Visible AI errors are lagging indicators.

Core form:

textScroll

visible AI errors are lagging indicators

Canonical sequence:

textScroll

H_AI↑ + Γ_AI drift↑ → O↓ → ε_AI spikes late

Plain meaning:

The first sign of AI failure is often not the hallucination, false refusal, unsafe compliance, public incident, or trust collapse. Failure often begins earlier as hidden classification drift, context erosion, routing distortion, feedback suppression, proxy divergence, correction burden, and repair load exported to users or downstream systems.

Pre-error debt form:

textScroll

Γ_AI drift + context_integrity↓ + FI↓ + repair_load↑ ⇒ ε_AI probability↑

Failure form:

textScroll

low visible AI error treated as safety proof ⇒ H_AI↑ + late incident

Primary variables:

ε_AI, H_AI, Γ_AI, classification_drift, context_integrity, routing_distortion, filtering_drift, interpretation_drift, proxy_divergence, correction_burden, user_burden_export, repair_load, incident_visibility, detection_lag, evaluation_lag, response_lag, recovery_lag, Au, Au_eff, FI, BΣ, R, R_eff, Φ_AI, L, Γ, Π, Ξ, ℛ, Θ, Σ, Ψ, Τ

Diagnostic signature:

Visible AI error remains low while confidence and deployment rise, classification trace falls, context integrity declines, user correction burden rises, feedback fails to reach the relevant AI layer, and repair load increases. This indicates AI error lag risk.

Failure risk:

AI error lag, AI latent failure, AI hidden error debt, AI classification drift, AI context erosion, AI routing distortion, AI proxy divergence, AI feedback suppression, AI correction load transfer, AI user burden export, AI incident shock, AI pseudo-safety, AI trust collapse, AI legibility collapse, hidden debt accumulation.

Restoration priority:

Audit AI error visibility, map hidden AI debt, classification drift, context erosion, routing distortion, proxy divergence, correction burden, and feedback reach; restore auditability, correction, appeal, and repair; reduce user-exported burden; and validate reduced AI error recurrence over time.

LAW-121 — AI as Γ-Amplifier Law

Registry

LAW-123 — AI U4 Truth Discipline Law